Hierarchical analysis of the impact of hospital characteristics on mortality in Alberta hospitals

1
Hierarchical analysis of the impact of hospital
characteristics on mortality in Alberta
hospitals

• Carole A. Estabrooks, RN, PhD
• William K. Midodzi, MSc
• Greta G. Cummings, RN, PhD(c)
• Kathryn L. Ricker, MSc
• Phyllis Giovannetti, RN, ScD

Sigma Theta Tau International November, 2003
2
Financial Support
Project Funding

• Alberta Heritage Foundation for Medical Research

Career support

• Canadian Institutes of Health Research (CIHR)
• Alberta Heritage Foundation for Medical Research

3
International Study of Hospital Outcomes
i n t r o d u c t i o n

• Purpose
• To determine the effects of the organization and
  nurse staffing of hospitals on patient and nurse
  outcomes.

4
Motivation for Research Agenda

• Widespread hospital restructuring and work
  redesign
• Changing hospital staffing patterns
• Absence of empirical evidence of these changes on
  outcomes

5
International samples
sample
Hospitals Nurses Alberta 109
6,526 British Columbia 97 2,838 England 32
5,006 Germany 30 4,000 Ontario 209 8,778
Scotland 7 5,238 U.S. (PA) 210
14,145
6
Alberta
sample
Hospitals Nurses 109 6,526 49
4,799 Criteria 5 nurses at least 20
beds
7
Sample Characteristics
Sample
(49 hospital) Population (109
hospital) Regular (FT/PT)
35.1/58.9
34.2/60.1 Casual
19.0
18.7 Female
97.2
97.5 Male
2.5
2.5 Age, yrs
40.4
40.9 Hours /wk 0-30
50.9
52.0 30 or more
46.5
45.5 Shift 8 Hrs
52.6
55.6 12 Hrs
35.8
33.8
Mixed 7.4
6.8 Diploma
(RN) 77.2
77.4 Baccalaureate
22.5
22.1

8
Data Sources
data sources

• Nurse survey
• Administrative data
• Alberta CIHI Hospital Inpatient Database
• Alberta Health Care Insurance Plan Registry
• Characteristics of Alberta acute care hospitals

9
The Alberta Nurse Survey

• Census of all staff nurses in hospitals
  (N12,345)
• Useable returns 6526 (52.8)

A Employment Characteristics B Nursing
Work Index (NWI) C Maslach Burnout
Inventory(MBI) D Job characteristics E
Last shift F Demographics G Site specific
questions
10
The Model
Organization Nurse Patient
11
30-day mortality model
Ability to develop relationships/Continuity of
care
Nursing training and skill variables
Quality of work environment

• Patient's characteristics
• Age
• Sex
• Co morbidity factors
• Complication
• Chronicity

30-day mortality
Other unknown determinants at the patient and
hospital levels

• Institutional factors
• Bed size
• Teaching hospital status
• Hospital location
• Hi-technology facility

12
Conceptual Basis for Hierarchical Modeling
Z
Level 2 (Hospital Level)
Level 1 (Patient Level)
X
Y
X Individual patient characteristics (age, sex,
admission diagnoses, comorbidity factors,
in-hospital complications, etc. ) Z Hospital
characteristics (bed size, location, teaching
status, nursing and physician factors, staffing,
etc.) Y The probability (or risk) of dying
within 30 days of admission to hospital
13
Nursing variables analyzed at the hospital level

• Nursing training and skill variables
• Nurse education level ( baccalaureate degree)
• Skill mix,
• Ability to develop a relationship with patients
• Job status casual or temporary staffs
• Perception of quality care
• Staffing, patients per nurse
• Patients care needs unattended
• Non-nursing activities performed by nurses
• Quality of work environment
• Nurse job satisfaction
• Support for non-floating policy
• Nurse autonomy
• Nurse-physician relationships
• Emotional abuse

14
Inter-hospital variation in risk adjusted 30-day
mortality
High mortality 16 hospitals
Low mortality 16 hospital
Average mortality 17 hospitals
Average co-morbidity score 0.0752
Average co-morbidity score 0.0725
Average co-morbidity score 0.0726
15
Nursing characteristics of study hospitals
16
The Linear Hierarchical Models(Outcome Risk
adjusted 30-day mortality)

• Model 1 (Patient Level) Controlling for patient
  factors
• Age, years
• Sex (Male/Female)

17
The Linear Hierarchical Models(Outcome Risk
adjusted 30-day mortality)

• Model 1 (Patient Level) Controlling for patient
  factors
• Age, years
• Sex (Male/Female)
• Model 2 (Hospital level) Institutional
  characteristics
• Bed size ( lt50, 51-150, gt150)
• Teaching status (gt1 resident per 4 bed, lt1
  residence per 4 bed, none)
• Location of hospitals (small cities pop. lt50K
  large cities pop. gt50K)

18
The Linear Hierarchical Models(Outcome Risk
adjusted 30-day mortality)

• Model 1 (Patient Level) Controlling for patient
  factors
• Age, years
• Gender (Male/Female)
• Model 2 (Hospital level) Instructional
  characteristics
• Bed size
• Teaching status Location of hospitals
• Model 3 (Hospital Level) Nursing related
  hospital characteristics
• Nurse education level
• Skill mix
• Job status casual or temporary
• Perception of quality care
• Staffing ratio last shift
• Patients care needs unattended
• Non-nursing activities performed by nurses
• Job satisfaction
• Support for non-floating policy
• Nurse autonomy
• Nurse-physician relationships
• Emotional abuse

19
Effect of nursing related hospital
characteristics
Constant - 5.19 - 0.50 x Nurse education
level - 0.27 x Skill Mix (RN to total nurse
staffs) 1.01 x Casual or temporary staffs -
0.14 x Perception of quality of care 0.01
x Staffing, pts. Per nurse ratio 0.08 x
Patients care needs unattended 0.04 x
Non-nursing activity performed - 0.12 x Job
satisfaction problems - 0.06 x Support for
non-floating policy - 0.22 x Nurse
autonomy - 0.11 x Nurse-physician
relationship 0.17 x Emotional
abuse Variance in 30-day mortality among
hospitals explained independently by
nursing-derived factors 33.1
Model adjusted for pts co-morbidity factors,
demographic variables and institutional factors
20
Percent of inter-hospital variation in30-day
mortality explained by each factors
Unknown determinants
Patients co-morbidity factors
Nursing-derived variables
Patients demographic variables
Institutional factors
21
Limitations

• Administrative data
• Aggregation
• Generalizability

22
Summary

• Lower patient mortality across hospitals was
  predicted in our models by

• higher nurse education levels
• work satisfaction
• quality of care
• support for non-floating policies
• nurse autonomy
• better nurse-physician relationships
• richer skill mix of nursing

23
Summary
Patients from hospitals with higher scores on the
following had significantly greater risk of dying
within 30-days of admission

• Higher percentage of casual employment
• Higher numbers of unmet patient care needs
• More non-nursing tasks completed
• Higher levels of reported emotional abuse
• Higher patient-to-nurse ratios

24
Implications